ChatGPT

Claims and evidence

• Released 30 November 2022Released as a free research preview. ● Primary source[1]page date; third paragraph

• Status Status unknownLaunched as a free research preview; OpenAI publishes no retirement date for the original November 2022 model, which had no API id, and the ChatGPT product has since moved to other models. ◌ Evidence incompleteNo source

• Derived from (fine tune) GPT-3.5OpenAI says ChatGPT is fine-tuned from a model in the GPT-3.5 series, using RLHF with the same methods as InstructGPT. ● Primary source[1]Methods

• Change · Other Trained for dialogue: human trainers wrote conversations playing both user and assistant, mixed with the InstructGPT dataset converted to dialogue format.Compared with GPT-3.5 ● Primary source[1]
• Change · Availability Offered as a free research preview on the web instead of through the API.Compared with GPT-3.5 ● Primary source[1]
• Change · Safety OpenAI says lessons from deploying GPT-3 and Codex informed its safety mitigations, including reductions in harmful and untruthful outputs through RLHF.Compared with GPT-3.5 ● Primary source[1]

• Parameters Not disclosedNot disclosed by OpenAI. No claim madeNo source
• Access consumer appFree web research preview at launch. ● Primary source[1]third paragraph

Lineage

Variants

No variants recorded in this record.

Related AI Radar coverage

AI Radar coverage starts in June 2026; no coverage linked yet.

All model releases from OpenAI on AI Radar →